Modeling Nonlinear Dynamic Systems Using Universal Learning Network with Filtering Mechanism

抄録

It has been already reported that the learning algorithm of Universal Learning Network (U.L.N.) by forward and backward propagation is useful for the modeling, managing, and controlling of the large scale complicated systems such as industrial plants, economic, social and life phenomena^[1][2][7][8].
Universal Learning Network is a network which can model and control naturally the large scale complicated systems and consists of nonlinearly operated nodes and multi-branches that may have arbitrary time delays including zero or minus ones. Therefore, U.L.N. can be applied to many kinds of systems which are difficult to be expressed as the ordinary first order difference equation with one sampling time delays.
In this paper, a new method is presented in order to make an optimal modeling of the dynamic system using U.L.N., where the word “optimal” means the establishment of the compact model as much as possible and the building of the model that has better performances. For the compactness of the modeling, a special filtering mechanism on all of the branches that cuts unnecessary branches are introduced. From simulation results, it has been clarified that selecting appropriate parameter variables in U.L.N. can make the compromised modeling in terms of modeling error and compactness of the model.